Passive fire protection · penetration seals

From the customer's enquiry to the owner's handover, every step checked against the tested configuration.

Gennemført 2.0 formalises the fire adviser's work. Enquiries are pre-checked against 678 tested rows from two ETAs while the customer types. Deviations go through a case workflow with precedents, four-eyes review and a hash-chained audit trail. Issued assessments, quotes, site records and handover packages can all be verified, file by file.

Open the app →Watch the 60-second tour
IFC4ETA-21/0110 + ETA-20/1322BCF 2.1MCPMIT
Live 3D register · fictitious demo building
0tested ETA rows, page and row cited
0answer key, hand-derived from the ETA pages
0browser = server, every check's text
0answered at intake, no adviser needed*
0incl. an end-to-end browser workflow
0in the SHA-256 manifest, re-hashed on every run

* simulation of 30 fictitious cases over 12 weeks.

Tour

Sixty seconds, end to end

Rendered frame by frame from the running app: no screen recording, no cuts in the 3D.

Workflow

Before and after

Same six-point deliverable as today: description, deviation, professional assessment, solution, assumptions, documentation. But each step is data that can be checked and reused.

Today

  1. Free-text formCustomer describes the task, attaches drawings.
  2. Follow-upAdviser calls for missing sizes and brands.
  3. PDF lookupTested set-ups found by hand, system by system.
  4. WordArgument written from scratch; precedents from memory.
  5. PDF sentNo link to the project, revisions or execution.

With Gennemført 2.0

  1. Structured enquiryFields per penetration, photo checklist, live pre-check.
  2. TriageTested set-ups answered at once; the rest routed east/west.
  3. Case + precedentNearest ETA row, deviation signature, earlier decisions.
  4. Decision + four eyesAdviser's own argument; reviewed by a colleague; hash-chained.
  5. Verifiable issuePDF with ETA excerpt, QR code, site checklist, owner package.
Modules

Nine modules, one knowledge base

M1Knowledge base across ETAs

Board seals and fire collars, 12 mounting situations, page images with the row highlighted, known source errors listed.

 M2Structured intake

Completeness score, photo checklist, live pre-check, routing and SLA.

 M3 · M4Cases, precedents, four eyes

Deviation signature to earlier decisions; author cannot review; SHA-256 chain.

 M5Quote with item numbers

Board, sealant, collar segments, clips, with the basis on every line.

 M6Execution and handover

Checklist from the ETA row, photo, signatures, inspector ≠ installer.

 M7 · M8Revisions and dashboard

Model revision diff by GlobalId; lead time, recurring deviations, regions.
Verification

Nothing has to be taken on trust

Every output has a check a person can do in minutes, and an automated one that runs on every change.

1The file

Every deliverable is in a SHA-256 manifest with the step, model, knowledge base and code that produced it. Drag a file onto Verificér; it is hashed in your browser, never uploaded.

2The answer

Every verdict shows the ETA page with the tested row highlighted, a link to the official PDF at that page, and each parameter as designed vs tested.

3The method

python -m gennemfoert.reproduce rebuilds everything from the source model and compares it with what is published: 9/9 checks.

File verification in the browser

Output-by-output verification table →

Gallery

Screens

Hub Receipt Tamper detection Dashboard
MCP

Ask it from Claude, Copilot or an IDE agent

Fourteen tools on the same engine: check a penetration, pre-check an enquiry, find precedents, read a case with its chain status, verify an assessment code, quote, revision impact.

{"mcpServers": {"gennemfoert": {
  "command": "python",
  "args": ["<repo>/src/mcp_server.py"]}}}